CN114707595B - Spark-based hyperspectral laser radar multichannel weighting system and method - Google Patents
Spark-based hyperspectral laser radar multichannel weighting system and method Download PDFInfo
- Publication number
- CN114707595B CN114707595B CN202210325224.4A CN202210325224A CN114707595B CN 114707595 B CN114707595 B CN 114707595B CN 202210325224 A CN202210325224 A CN 202210325224A CN 114707595 B CN114707595 B CN 114707595B
- Authority
- CN
- China
- Prior art keywords
- channel
- data
- pulse
- laser radar
- echo pulse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000003595 spectral effect Effects 0.000 claims abstract description 67
- 238000012545 processing Methods 0.000 claims abstract description 39
- 238000013500 data storage Methods 0.000 claims abstract description 22
- 238000009825 accumulation Methods 0.000 claims abstract description 18
- 238000012937 correction Methods 0.000 claims abstract description 14
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 238000005457 optimization Methods 0.000 claims abstract description 5
- 238000001228 spectrum Methods 0.000 claims description 50
- 101000932776 Homo sapiens Uncharacterized protein C1orf115 Proteins 0.000 claims description 13
- 102100025480 Uncharacterized protein C1orf115 Human genes 0.000 claims description 13
- 239000000203 mixture Substances 0.000 claims description 12
- 238000010801 machine learning Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000003786 synthesis reaction Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 230000004382 visual function Effects 0.000 claims description 2
- 230000000875 corresponding effect Effects 0.000 description 42
- 230000006870 function Effects 0.000 description 10
- 238000000354 decomposition reaction Methods 0.000 description 6
- 238000003672 processing method Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000007792 addition Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000002592 echocardiography Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002386 leaching Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/18—File system types
- G06F16/182—Distributed file systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Optical Radar Systems And Details Thereof (AREA)
Abstract
The invention discloses a Spark-based hyperspectral lidar multichannel weighting system, which comprises a data distribution layer, a data storage layer, a hyperspectral lidar multichannel weighting method and a data processing layer, wherein the data storage layer comprises a batch processing layer, a service layer and a speed layer; performing time domain correction on the smoothed channel data, and outputting the channel data after the time domain correction; carrying out weighted accumulation on the time domain corrected channel data, and outputting the weighted accumulated channel data; and carrying out parameter extraction optimization on the weighted and accumulated channel data, and outputting a distance data set and an intensity data set. The method effectively improves the extraction precision of the spectral information and the distance information of each channel and each pulse of the hyperspectral laser radar waveform data and the running speed of the hyperspectral laser radar multichannel weighting processing algorithm.
Description
Technical Field
The invention relates to the technical field of remote sensing mapping, in particular to a Spark-based hyperspectral lidar multichannel weighting system, and also relates to a Spark-based hyperspectral lidar multichannel weighting method, which is suitable for high-speed real-time and offline waveform data of a plurality of channels of hyperspectral lidar.
Background
The hyperspectral laser radar is used as a brand new earth observation technology, a supercontinuum laser is adopted, and spectrum information and distance information of different wave bands can be obtained on one laser footprint, so that distance detection and target identification of a plurality of ground features are realized. Compared with the traditional single-wavelength laser radar, the hyperspectral laser radar has more channels, adopts multiple types of detector arrays to cover wide-spectrum detection (400 nm-2500 nm) from visible light to near infrared, and has more abundant acquired ground object target information. Because the energy distribution of the super-continuum spectrum laser is uneven and the reflection characteristics of different ground objects are different; the response curves, detection efficiency and the like of the multiple types of detectors are different, so that the backward scattering echo waveforms, strengths and the like of different channels are obviously different; the multi-channel waveform data of the hyperspectral laser radar has the characteristics of massive, high-speed and other large data, and is hundreds or even thousands of times of the existing single-channel single-waveform data. The above challenges the extraction accuracy and extraction speed of the multichannel spectral waveform parameters. The traditional single-channel single-wavelength decomposition algorithm only carries out single-thread decomposition on single channel and single waveform, and is not applicable to the decomposition of hyperspectral laser radar multi-pass mass waveform data. The invention provides a Spark-based hyperspectral lidar multichannel weighting system and a Spark-based hyperspectral lidar multichannel weighting method. The method can improve the extraction precision of the spectral information and the distance information of each channel and each pulse of the hyperspectral laser radar waveform data and the operation rate of a hyperspectral laser radar mass multichannel data weighting processing algorithm.
Disclosure of Invention
The invention aims to solve the problems in the prior art, provide a Spark-based hyperspectral laser radar multichannel weighting system and a Spark-based hyperspectral laser radar multichannel weighting method, and effectively improve the extraction precision of spectral information and distance information of each channel of hyperspectral laser radar waveform data and the operation rate of a hyperspectral laser radar multichannel weighting processing algorithm by establishing a Spark-based hyperspectral laser radar massive waveform data multichannel weighting algorithm system and a Spark calculation analysis frame by utilizing a hyperspectral laser radar multichannel weighting processing algorithm.
The Spark-based hyperspectral lidar multichannel weighting system comprises a data distribution layer, a data storage layer,
the data distribution layer is interacted with the hyperspectral laser radar acquisition system and is used for acquiring the channel data of a large number of multi-channel full waveforms of the hyperspectral laser radar system, the channel data of the large number of multi-channel full waveforms of the hyperspectral laser radar system are pre-classified according to different spectrum channels and are sent to corresponding data storage layer nodes of the data storage layer, and the search result is timely fed back to a user;
the data storage layer includes a batch processing layer, a service layer, and a speed layer,
in the indexing stage of the channel data of the mass multi-channel full waveform of the off-line hyperspectral laser radar system:
the mass processing layer is used for reading mass channel data of the multi-channel full waveform of the hyperspectral laser radar from the HDFS distributed file system, carrying out distributed parallel preprocessing on the read mass channel data of the multi-channel full waveform of the hyperspectral laser radar to generate a mass noise data table and a mass waveform data table, carrying out secondary distributed parallel processing on the mass waveform data table to generate corresponding channel data of each spectrum channel, and storing the corresponding channel data in the HDFS distributed file system;
in the channel data indexing stage of a mass multi-channel full waveform of a real-time hyperspectral laser radar system:
the service layer is used for reading the channel data of the mass real-time hyperspectral laser radar multichannel full waveform from the HDFS distributed file system, carrying out distributed parallel preprocessing on the read channel data of the mass hyperspectral laser radar multichannel full waveform to generate a mass noise data table and a mass waveform data table, carrying out secondary distributed parallel processing on the mass waveform data table to generate corresponding channel data of each spectrum channel, storing the channel data in the HDFS distributed file system,
the speed layer adopts a SparkStreaming streaming processing framework, and a combined query result of the service layer and the speed layer is directly returned through real-time query and is stored in an HDFS distributed file system.
The Spark-based hyperspectral laser radar multichannel weighting system further comprises a data processing layer, wherein the data processing layer is used for decoding a massive noise data table of the data storage layer and channel data corresponding to each channel, the Spark framework based on the hyperspectral laser radar massive waveform data multichannel weighting processing method is used for analyzing the channel data by using the Spark-based hyperspectral laser radar multichannel weighting method, and the output distance and intensity data set RDD1 is stored in the HDFS distributed file system; combining a point cloud synthesis algorithm to generate a hyperspectral laser radar point cloud data set RDD2, and storing the data set RDD2 in an HDFS distributed file system; and performing cluster learning training on the first k data in the hyperspectral laser radar point cloud data set RDD2, outputting a machine learning data set RDD3, and storing the machine learning data set RDD3 in an HDFS distributed file system.
The Spark-based hyperspectral lidar multichannel weighting system further comprises a data application layer, wherein the data application layer is used for receiving a distance and intensity data set RDD1, a hyperspectral lidar point cloud data set RDD2 and a machine learning data set RDD3 distributed by the data application layer and setting up a corresponding visual function window.
A Spark-based hyperspectral lidar multichannel weighting method comprises the following steps:
step 1, carrying out smooth filtering on channel data distributed by a manager, and outputting the smoothed channel data;
step 2, performing time domain correction on the smoothed channel data, and outputting the time domain corrected channel data;
step 3, carrying out weighted accumulation on the time-domain corrected channel data, and outputting the weighted accumulated channel data;
and 4, carrying out parameter extraction optimization on the weighted and accumulated channel data, and outputting a distance data set and an intensity data set.
The time domain correction in step 2 as described above is based on the following formula:
T j,i ＝(t 2,j,i -t j,i ”)-(t 1,j -t j ')
wherein T is j,i Representing the time of flight of the ith backscatter echo pulse of the jth spectral channel after time domain correction; t is t 2，j,i Representing the instant t of the recorded ith backscattered echo pulse of the jth spectral channel j,i "represents the propagation time error, t, of the ith backscattered echo pulse of the jth spectral channel 1，j Representing the recorded moment of the pulse emitted by the jth spectral channel, t j ' represents the propagation time error of the j-th spectral channel transmit pulse.
Step 3 as described above comprises the steps of:
step 3.1, calculating a time-domain corrected channel data multi-echo quality value MEQ based on the following formula j ，
In f (x) j Represents time-domain corrected channel data including transmit pulse and back-scattered echo pulse data, m represents waveform length, S noise,j Representing the mean square error of background noise of the jth spectral channel;
step 3.2, establishing a multi-channel weighted accumulation model f a (x) The weighted accumulated channel data ch3_index (j) is output,
wherein omega is j A weight representing the jth spectral channel; n represents the number of spectral channels participating in the accumulation; s is S noise，j Representing the mean square error of the jth spectral channel noise; f (f) a (x) Weighting the accumulation model for multiple channels; f (x) j Representing time domain corrected channel data, including transmit pulse and backscattered echo pulse data.
Step 4 as described above comprises the steps of:
step 4.1, establishing a Gaussian mixture fitting model;
step 4.2, extracting a to-be-optimized transmitting pulse parameter and a to-be-optimized back scattering echo pulse parameter from the weighted and accumulated channel data CH3_index (j), and respectively optimizing the to-be-optimized transmitting pulse parameter and the to-be-optimized back scattering echo pulse parameter according to a Gaussian mixture fitting model to obtain an optimal transmitting pulse parameter and an optimal back scattering echo pulse parameter;
and 4.3, calculating the distance between the central position of the optimal back scattering echo pulse and the central position of the corresponding transmitting pulse according to the optimal transmitting pulse parameter and the optimal back scattering echo pulse parameter.
Step 4.2 as described above comprises the steps of:
step 4.2.1 searching the weighted and accumulated channel data CH3_index (j) for a maximum value ψ of the transmit pulse intensity for each spectral channel 0,j The method comprises the steps of carrying out a first treatment on the surface of the Obtaining the maximum value psi of the emission pulse of each spectral channel 0,j Corresponding center position X 0,j The method comprises the steps of carrying out a first treatment on the surface of the Acquiring emission pulse intensity maxima ψ 0,j Inflection point spacing F on left and right sides 0,j Maximum value psi of emission pulse intensity of each spectral channel 0,j Center position X 0,j Inflection point distance F 0,j Forming a group of to-be-optimized transmitting pulse parameters;
step 4.2.2, sequentially inputting each group of emission pulse parameters to be optimized into a mixed Gaussian fitting model, and adopting LM calculation based on nonlinear least square curve fittingObtaining the optimal emission pulse parameters corresponding to each spectrum channel by a method comprising the maximum value psi 'of the optimal emission pulse intensity' 0,j Optimum center position X' 0,j Inflection point spacing F' 0,j ；
Step 4.2.3 searching the weighted accumulated channel data ch3_index (j) for a maximum value ψ of the respective backward scattered echo pulse intensities for each spectral channel j，i The method comprises the steps of carrying out a first treatment on the surface of the Obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Corresponding center position X j,i The method comprises the steps of carrying out a first treatment on the surface of the Obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Inflection point spacing F on left and right sides j,i The maximum value psi of the back scattering echo pulse intensity corresponding to each back scattering echo pulse corresponding to the spectrum channel j，i Center position X j,i Inflection point distance F j,i Forming a group of back scattering echo pulse parameters to be optimized;
step 4.2.4, synchronously traversing each spectrum channel in a multithread way, inputting each group of backscattering echo pulse parameters to be optimized in the spectrum channel into a mixed Gaussian fitting model, obtaining each group of optimized backscattering echo pulse parameters corresponding to the spectrum channel by adopting an LM algorithm based on nonlinear least squares curve fitting, and in each group of optimized backscattering echo pulse parameters corresponding to the spectrum channel, enabling the amplitude of the backscattering echo pulse to be smaller than a threshold thr j Deleting the corresponding optimized backscattering echo pulse parameters to obtain the optimal backscattering echo pulse parameters including the maximum value psi 'of the backscattering echo pulse intensity' j，p Center position X' j,p Inflection point spacing F' j,p 。
The mixture gaussian fitting model as described above is:
wherein f j (t) is a mixed Gaussian fitting model of the jth spectrum channel, and ψ represents the amplitude; t represents a sampling time; x represents a central position; f represents half width; s is S noise,j Representing the noise mean square error of the jth spectral channel.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention carries out system operation on massive multi-channel full-waveform data of the hyperspectral laser radar by distributing the data to the data application, builds a rapid query processing system, solves the problem of low efficiency of the massive multi-channel waveform data of the hyperspectral laser radar in an index operation stage, and has good systematicness and expandability.
2. According to the invention, the Spark frame is combined with the channel data of the hyperspectral laser radar mass multichannel waveform, so that a Spark parallel computing system of the hyperspectral laser radar multichannel weighting method based on Spark is formed, and compared with the traditional single-thread algorithm, the operation speed of the hyperspectral laser radar mass waveform data multichannel weighting algorithm based on Spark is effectively improved, and the operation speed of the hyperspectral laser radar mass waveform data multichannel weighting processing algorithm is greatly improved.
3. The method adopts a Spark-based hyperspectral laser radar multichannel weighting method to establish a hyperspectral laser radar multichannel waveform channel data time domain correction method, so that channel data with deviation of flight time is effectively subjected to time domain unified correction and integrated into subsequent data processing; a multi-channel weighted accumulation model is established, and waveform pulse signals of a strong channel are utilized to position waveform pulse signals of a weak channel, so that the problems that weak echoes cannot be detected and the initialization precision of waveform parameters is low are solved compared with the traditional single-channel waveform decomposition algorithm, and the precision of distance detection, ground object classification recognition and the like is greatly improved.
Detailed Description
The invention will be further illustrated with reference to the following examples, which are set forth in order to facilitate a person skilled in the art in the understanding and practice of the invention, it being understood that the examples described herein are for the purpose of illustration and explanation only and are not intended to limit the invention to that which is described. The invention is capable of other and different embodiments and of being practiced or being modified in various ways without departing from the spirit of the invention.
The multichannel weighting processing system based on hyperspectral laser radar mass waveform data comprises a data distribution layer, a data storage layer, a data processing layer and a data application layer, and is characterized in that:
the data distribution layer is interacted with the hyperspectral laser radar acquisition system and is used for acquiring the channel data of a large number of multi-channel full waveforms of the hyperspectral laser radar system, the channel data of the large number of multi-channel full waveforms of the hyperspectral laser radar system are pre-classified according to different spectrum channels and are sent to corresponding data storage layer nodes of the data storage layer, and the search result is timely fed back to a user;
the data storage layer is connected with the data distribution layer and comprises a batch processing layer, a service layer and a speed layer.
(1) And starting a batch processing layer in an index stage of the channel data of the mass multi-channel full waveform of the off-line hyperspectral laser radar system, and reading the channel data of the mass hyperspectral laser radar multi-channel full waveform from the HDFS distributed file system, and carrying out distributed parallel preprocessing on the read channel data of the mass hyperspectral laser radar multi-channel full waveform. In order to pick the transmit pulses and the corresponding backscattered echo pulses in the channel data of the hyperspectral lidar multichannel massive waveforms, an evaluation of the noise level is required. It is considered that the background noise satisfies three times the standard deviation of the average background noise, and thus a noise threshold thr is set to distinguish the signal from the noise. Since the background noise follows a normal distribution, no signal echo is present before the transmit pulse, the noise is generally chosen to be the length of time from the start of the transmit pulse to calculate its average value m noise And mean square error S noise Further, a noise threshold thr is calculated.
thr j ＝3S noise,j +m noise,j (1)
Wherein thr j Is the noise threshold of the jth spectral channel to distinguish between signal and noise; s is S noise，j Mean square error, m representing noise of jth spectral channel noise Representing the average value of the noise of the jth spectral channel.
When hyperspectral laser radar is used in seaThe amplitude of the channel data of the magnitude waveform is less than thr j Considered as Noise data, and stored in a massive Noise data table noise_rdd;
when the amplitude of channel data of mass waveforms of the hyperspectral laser radar is larger than a threshold thr j The effective waveform data is considered to be stored in the mass waveform data table data_rdd. Generating a massive Noise data table noise_RDD and a massive waveform data table data_RDD, performing secondary distributed parallel processing on the massive waveform data table data_RDD, generating channel data CH_RDD (j) corresponding to each spectrum channel, and storing the channel data CH_RDD in an HDFS distributed file system;
(2) In a channel data indexing stage of a mass multi-channel full waveform of a real-time hyperspectral laser radar system, a service layer is started and used for reading mass channel data of the mass real-time hyperspectral laser radar multi-channel full waveform from an HDFS distributed file system, carrying out distributed parallel preprocessing on the read mass channel data of the mass hyperspectral laser radar multi-channel full waveform to generate a mass Noise data table noise_RDD and a mass waveform data table data_RDD, carrying out secondary distributed parallel processing on the mass waveform data table data_RDD to generate channel data CH_RDD (j) corresponding to each spectrum channel, and storing the channel data CH_RDD (j) in the HDFS distributed file system; in the speed layer, a SparkStreaming streaming processing framework can be adopted, and along with the increase of the data volume, the combined query results of the service layer and the speed layer are directly returned through real-time query and stored in an HDFS distributed file system;
the HDFS distributed file system is used for storing channel data of the multi-channel full waveform of the high-intensity spectrum laser radar in a distributed manner and various data tables and final data sets generated in all intermediate steps of a multi-channel weighting processing method of the high-intensity spectrum laser radar mass waveform data based on Spark;
the batch processing layer is used for preprocessing the off-line mass multi-channel full-waveform channel data of the hyperspectral laser radar, constructing a Noise index function corresponding to query on the mass channel data of the hyperspectral laser radar of all channels, generating a corresponding mass Noise data table noise_RDD, a mass waveform data table data_RDD and channel data CH_index (j), and storing the corresponding mass Noise data table noise_RDD, the mass waveform data table data_RDD and the channel data CH_index (j) in the HDFS distributed file system.
The service layer is used for supplementing the batch processing layer in real time, is used for preprocessing the real-time mass multi-channel full-waveform channel data, constructs a Noise index function corresponding to query on the hyperspectral laser radar mass data of all spectrum channels, a waveform data index function and a channel data index function, generates a corresponding real-time mass Noise data table noise_RDD, a mass waveform data table data_RDD and channel data CH_index (j), and stores the corresponding real-time mass Noise data table noise_RDD, the mass waveform data table data_RDD and the channel data CH_index (j) in the HDFS distributed file system.
And the speed layer adopts a SparkStreaming streaming processing framework, and as channel data increases, a result is directly returned through real-time query, so that a combined query result of the service layer and the speed layer is obtained.
The data processing layer is connected with the data storage layer and is used for decoding a mass Noise data table noise_RDD of the data storage layer and channel data CH_index (j) corresponding to each channel, analyzing channel data by utilizing a multi-channel weighting processing method based on mass waveform data of the hyperspectral laser radar based on a Spark frame of the multi-channel weighting processing method of mass waveform data of the hyperspectral laser radar, outputting a distance and intensity data set RDD1 and storing the distance and intensity data set RDD1 in an HDFS distributed file system; generating a hyperspectral laser radar point cloud data set RDD2 by combining a point cloud synthesis algorithm, and storing the data set RDD2 in an HDFS distributed file system; performing cluster learning training on the first k data in the hyperspectral laser radar point cloud data set RDD2 by using a machine learning correlation algorithm, outputting a machine learning data set RDD3, and storing the machine learning data set RDD3 in an HDFS distributed file system;
the data application layer is connected with the data processing layer and is used for receiving a distance and intensity data set RDD1, a hyperspectral laser radar point cloud data set RDD2 and a machine learning data set RDD3 distributed by the data application layer, correspondingly setting application function windows such as a space position measurement function module, a ground feature identification monitoring function module and a ground feature classification function module, and providing a visual interface for a user.
The Spark framework based on the hyperspectral lidar massive waveform data multichannel weighting processing algorithm of the embodiment is provided as follows:
the driver is used for running main functions, creating and starting Spark context, building an operation environment required by a Spark application program, reading index tables noise_RDD and CH_index (j) of the hyperspectral laser radar multichannel data storage layer from the HDFS distributed file system, and distributing the index tables noise_RDD and CH_index (j) to the executor in parallel;
the manager is connected with the driver and used for controlling the whole Spark cluster and monitoring all workstations; the SparkContext in the driver applies for registration to the manager and applies for running the executor resource; and returning a data set index table and a distance and intensity data set RDD1 generated in the middle link of the Spark frame of the mass waveform data multi-channel weighting processing method based on the hyperspectral laser radar, and storing the data set index table and the distance and intensity data set RDD1 in an HDFS distributed file system.
The executor is connected with the manager and is used for receiving and processing mass Noise data tables noise_RDD and channel data CH_index (j) of the hyperspectral laser radar multichannel data storage layer distributed by the manager; distributing each index table to a distributed parallel workstation, and executing corresponding work tasks by utilizing a hyperspectral laser radar multichannel weighting processing algorithm and adopting multithreading streaming parallel calculation; the intermediate results of all the distributed workstations and the distance and intensity data set RDD1 are summarized and submitted to a manager, and after the operation is finished, the intermediate results are stored in an HDFS distributed file system, and the sparkContext logs off the manager and releases all resources;
a Spark-based hyperspectral lidar multichannel weighting method comprises the following steps:
step 1, waveform preprocessing, specifically:
smooth filtering is carried out on channel data distributed by a manager, the channel data comprise offline channel data CH_index (j) and a real-time channel data table CH_index (j), and the offline channel data CH_index (j) and the real-time channel data table CH_index (j) both comprise transmitting pulses and backward scattering echo pulses, and specifically:
and leaching each spectrum channel by using a low-pass filter, filtering noise influence of each spectrum channel, and outputting smoothed channel data.
Step 2, performing time domain correction on the smoothed channel data, and outputting the time domain corrected channel data, wherein the time domain corrected channel data specifically comprises the following steps:
due to factors such as response time deviation of different detectors, propagation deviation of a plurality of spectrum channels in the instrument and the like, the back scattering echo pulse of the hyperspectral laser radar has flight time difference among a plurality of different spectrum channels, and the time domain unified correction and calibration are required to be carried out on the back scattering echo pulse flight time of the plurality of spectrum channels.
The wave propagation time of the same target in different spectrum channels has a time difference, the receiving time of the target echo is equal to the difference between the recording time of the target echo and the propagation time error of the target echo, the receiving time of the transmitting pulse is equal to the difference between the recording time of the transmitting pulse and the propagation time error of the transmitting pulse, and the propagation time of the target echo is based on the transmitting pulse receiving time of each spectrum channel. After the echo waveform time of each spectrum channel is synchronized, the time is uniformly calibrated, so that the position distribution of all channels can be integrated into the subsequent data processing,
T j,i ＝(t 2,j,i -t j,i ”)-(t 1,j -t j ') (2)
in the formula (2), T j,i Representing the time of flight of the ith backscatter echo pulse of the jth spectral channel after time domain correction; t is t 2，j,i Representing the recorded ith backscattering echo pulse of the jth spectral channel, typically referenced to the time at the center of the backscattering echo pulse; t is t j,i "represents the propagation time error of the ith backscattered echo pulse of the jth spectral channel; t is t 1，j Representing the recorded moment when the j-th spectral channel emits a pulse, generally referenced to the moment at the center of the emitted pulse; t is t j ' represents the propagation time error of the j-th spectral channel transmit pulse.
Step 3, carrying out weighted accumulation on the time-domain corrected channel data, and outputting the weighted accumulated channel data, wherein the weighted accumulation is specifically as follows:
step 31, calculating a multi-echo quality value MEQ of channel data after time domain correction j (Multiple echo quality)，
In formula (3), MEQ j Representing the multi-echo quality, f (x), of the jth spectral channel j Represents time domain corrected channel data, including transmit pulse and back scatter echo pulse data, and m represents waveform length. S is S noise,j Representing the mean square error of the background noise for the jth spectral channel. Multi-echo quality value MEQ j The multi-channel multi-echo data quality of the hyperspectral laser radar is positively correlated with the multi-echo quality, and the quality of the multi-channel massive multi-echo data of the hyperspectral laser radar can be well represented.
Step 3.2, establishing a multi-channel weighted accumulation model f a (x)，
Because the hyperspectral laser radar system has more channels, the signals of the transmitting pulse and the back scattering echo pulse of some channels are weaker, the waveform decomposition is carried out on a single channel, the influence of noise is caused, and the initialized waveform parameter set has deviation. Establishing a multi-channel weighted accumulation model f by utilizing the interconnectivity between channels a (x) The strong signal channel can be utilized to precisely position the weak signal channel, the waveform parameter extraction precision is improved,
in the formulae (4) - (5), ω j A weight representing the jth spectral channel; n represents the number of spectral channels participating in the accumulation; s is S noise，j Representing the mean square error of the jth spectral channel noise; f (f) a (x) Storing in channel data ch3_index (j) for a multi-channel weighted accumulation model; f (x) j Represents time-domain corrected channel data ch2_index (j) containing transmit pulse and backward-scattered echo pulse data.
If there is only one target in the laser footprint, MEQ according to each channel j Value, weighting cumulative model f to multiple channels a (x) Adding f (x) in descending order of the middle j The method comprises the steps of carrying out a first treatment on the surface of the If there are multiple targets in the laser footprint, the weight ω of each spectral channel is calculated according to equation (4) j Two cases of weighted accumulation are proposed: first, weight ω of each spectral channel j Are identical, MEQ according to each channel j Value, weighting cumulative model f to multiple channels a (x) Adding f (x) in descending order j The weighted accumulation method is suitable for the condition of lower noise level, and the data quality of each spectrum channel is similar, so that the calculation efficiency is higher; second, weight ω of each spectral channel j In inverse proportion to the corresponding background noise level, the spectral channels are weighted and accumulated sequentially according to equation (5), which is more advantageous to suppress the influence of the spectral channels with high noise level on the accumulated channel data ch3_index (j). For each addition of time-domain corrected channel data of one spectral channel to channel data ch2_index (j), the multi-echo quality MEQ of the newly added spectral channel is updated synchronously according to equation (3) j Multiple echo quality MEQ values up to newly added spectral channels j The value is improved or all spectral channels participate in the operation of equation (5), the updating is stopped and the weighted accumulated channel data ch3_index (j) is output.
Step 4, performing parameter extraction optimization on the weighted and accumulated channel data ch3_index (j), and outputting a distance data set and an intensity data set RDD1, which specifically are:
step 4.1, establishing a Gaussian mixture fitting model,
because the hyperspectral laser radar pulse signals approximate Gaussian distribution, when the same laser beam scans the ground object, light spots can be captured by different targets, and a plurality of echo pulses can appear in a back scattering echo waveform, and the back scattering echo waveform can be regarded as linear superposition of a plurality of Gaussian functions. The conventional gaussian decomposition model does not take into account the state of multiple echoes, which may lead to errors in the initialization of waveform parameters. The Gaussian mixture fitting model is established, and the extraction precision of mass waveform parameters of the multi-channel hyperspectral laser radar can be effectively improved aiming at the back-scattering multi-echo channel.
Wherein f j (t) is a mixed Gaussian fitting model of the jth spectrum channel, and represents waveform data of the jth spectrum channel after fitting; k represents the number of transmit pulses or the number of backscattered echo pulses of the spectral channel; psi represents amplitude; t is the sampling time; x represents a central position; f represents half width; s is S noise,j Representing the noise mean square error of the jth spectral channel.
And 4.2, extracting a to-be-optimized transmitting pulse parameter and a to-be-optimized back scattering echo pulse parameter from the weighted and accumulated channel data CH3_index (j), respectively optimizing the to-be-optimized transmitting pulse parameter and the to-be-optimized back scattering echo pulse parameter according to a Gaussian mixture fitting model, obtaining an optimal transmitting pulse parameter and an optimal back scattering echo pulse parameter, and storing the optimal transmitting pulse parameter and the optimal back scattering echo pulse parameter into a parameter set RDD 0.
Step 4.2.1 searching the weighted and accumulated channel data CH3_index (j) for a maximum value ψ of the transmit pulse intensity for each spectral channel 0,j The method comprises the steps of carrying out a first treatment on the surface of the Obtaining the maximum value psi of the emission pulse of each spectral channel 0,j Corresponding center position X 0,j The method comprises the steps of carrying out a first treatment on the surface of the Searching the maximum value psi of the transmitted pulse in each spectral channel 0,j Corresponding inflection point is used for obtaining the maximum value psi of the intensity of the transmitting pulse 0,j Inflection point spacing F on left and right sides 0,j Maximum value psi of emission pulse intensity of each spectral channel 0,j Center position X 0,j Inflection point distance F 0,j Forming a group of emission pulse parameters to be optimized;
step 4.2.2, sequentially inputting each group of emission pulse parameters to be optimized into a Gaussian mixture fitting model, namely psi 0,j Substituted into psi, X 0,j Substituted into X, F 0,j Substituting into F, obtaining the optimal transmitting pulse parameters corresponding to each spectrum channel by adopting LM algorithm based on nonlinear least square curve fitting, wherein the optimal transmitting pulse parameters comprise the maximum value psi 'of the optimal transmitting pulse intensity' 0,j Optimum center position X' 0,j Inflection point spacing F' 0,j And stored in the optimization parameter set rdd0_1;
step 4.2.3 searching the weighted accumulated channel data ch3_index (j) for a maximum value ψ of the respective backward scattered echo pulse intensities for each spectral channel j，i The method comprises the steps of carrying out a first treatment on the surface of the Obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Corresponding center position X j,i The method comprises the steps of carrying out a first treatment on the surface of the Searching for a maximum value ψ of the respective backscattered echo pulse intensities for each spectral channel j，i A corresponding inflection point; obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Inflection point spacing F on left and right sides j,i The maximum value psi of the back scattering echo pulse intensity corresponding to each back scattering echo pulse corresponding to the spectrum channel j，i Center position X j,i Inflection point distance F j,i Forming a group of to-be-optimized back-scattering echo pulse parameters, wherein each spectrum channel corresponds to i groups of to-be-optimized back-scattering echo pulse parameters;
step 4.2.4, synchronously traversing each spectrum channel in a multithread way, and inputting each group of back scattering echo pulse parameters to be optimized in the spectrum channel into a Gaussian mixture fitting model, namely psi j，i Substituted into psi, X j,i Substituted into X, F j,i Substituting into F, obtaining each group of optimized scattered echo pulse parameters corresponding to the spectrum channel by adopting LM algorithm based on nonlinear least square curve fitting, and enabling the amplitude of the backscattered echo pulse to be smaller than a threshold thr in each group of optimized scattered echo pulse parameters corresponding to the spectrum channel j Deleting the corresponding optimized back scattering echo pulse parameters to obtain the optimal back scattering echo pulse parameters, wherein the optimal back scattering echo pulse parameters comprise the maximum value psi 'of the back scattering echo pulse intensity' j，p Center position X' j,p Inflection point spacing F' j,p Storing each optimal back scattering echo pulse parameter corresponding to the spectrum channel in an optimal parameter set RDD0_1;
step 4.3, calculating the distance of the target,
the center position X 'of the optimal backscattered echo pulse in the optimized parameter set RDD0_1' j,p And the center position X 'of the transmit pulse in the corresponding optimum transmit pulse parameters' 0,j Inputting (7), calculating to obtain the central position X 'of the optimal back scattering echo pulse' j,p And the center position X 'of the corresponding transmitted pulse' 0,j Distance L between j,p A distance and intensity data set RDD1 is output,
L j,p ＝(X' j,p -X' 0,j )×c÷2 (7)
wherein L is j,p Representing the center position X 'of the optimal backscattered echo pulse' j,p And the center position X 'of the corresponding transmitted pulse' 0,j A distance therebetween; c represents the speed of light. Maximum value psi 'of optimal emission pulse intensity' 0,j Maximum value psi 'of optimal back scattering echo pulse intensity' j，p Distance L of target j,p Sequentially into the distance and intensity data set RDD 1.
It should be noted that the specific embodiments described in this application are merely illustrative of the spirit of the invention. Those skilled in the art may make various modifications or additions to the described embodiments or substitutions in a similar manner without departing from the spirit of the invention or exceeding the scope of the invention as defined in the accompanying claims.
Claims (2)
1. The Spark-based hyperspectral lidar multichannel weighting system comprises a data distribution layer, and is characterized by also comprising a data storage layer,
the data distribution layer is interacted with the hyperspectral laser radar acquisition system and is used for acquiring the massive multi-channel full-waveform channel data of the hyperspectral laser radar system, the massive multi-channel full-waveform channel data of the hyperspectral laser radar system are pre-classified according to different spectrum channels and sent to corresponding data storage layer nodes of the data storage layer, and the search result is timely fed back to a user;
the data storage layer includes a batch processing layer, a service layer, and a speed layer,
in the indexing stage of the channel data of the mass multi-channel full waveform of the off-line hyperspectral laser radar system:
the mass processing layer is used for reading mass channel data of the multi-channel full waveform of the hyperspectral laser radar from the HDFS distributed file system, carrying out distributed parallel preprocessing on the read mass channel data of the multi-channel full waveform of the hyperspectral laser radar to generate a mass noise data table and a mass waveform data table, carrying out secondary distributed parallel processing on the mass waveform data table to generate corresponding channel data of each spectrum channel, and storing the corresponding channel data in the HDFS distributed file system;
in the channel data indexing stage of a mass multi-channel full waveform of a real-time hyperspectral laser radar system:
the service layer is used for reading massive real-time multi-channel full-waveform channel data of the hyperspectral laser radar from the HDFS distributed file system, carrying out distributed parallel preprocessing on the read real-time massive multi-channel full-waveform channel data of the hyperspectral laser radar to generate a massive noise data table and a massive waveform data table, carrying out secondary distributed parallel processing on the massive waveform data table to generate corresponding channel data of each spectrum channel, storing the corresponding channel data in the HDFS distributed file system,
the speed layer adopts a SparkStreaming streaming processing framework, the combined query results of the service layer and the speed layer are directly returned through real-time query and are stored in an HDFS distributed file system,
the device also comprises a data processing layer, a high-speed data storage layer and a data storage layer, wherein the data processing layer is used for decoding a mass noise data table of the data storage layer and channel data corresponding to each channel, analyzing the channel data by using a Spark-based high-spectrum laser radar multichannel weighting method based on a Spark-based high-spectrum laser radar multichannel weighting method, outputting a distance and intensity data set RDD1 and storing the distance and intensity data set RDD1 in an HDFS distributed file system; combining a point cloud synthesis algorithm to generate a hyperspectral laser radar point cloud data set RDD2, and storing the data set RDD2 in an HDFS distributed file system; cluster learning training is carried out on the first k data in the hyperspectral laser radar point cloud data set RDD2, a machine learning data set RDD3 is output, the data are stored in an HDFS distributed file system,
the system further comprises a data application layer, wherein the data application layer is used for receiving a distance and intensity data set RDD1, a hyperspectral laser radar point cloud data set RDD2 and a machine learning data set RDD3 distributed by the data application layer and setting up a corresponding visual function window.
2. The Spark-based hyperspectral lidar multichannel weighting method is characterized by comprising the following steps of:
step 1, carrying out smooth filtering on channel data distributed by a manager, and outputting the smoothed channel data;
step 2, performing time domain correction on the smoothed channel data, and outputting the time domain corrected channel data;
step 3, carrying out weighted accumulation on the time-domain corrected channel data, and outputting the weighted accumulated channel data;
step 4, parameter extraction optimization is carried out on the weighted and accumulated channel data, a distance data set and an intensity data set are output,
the time domain correction in step 2 is based on the following formula:
T j,i ＝(t 2,j,i -t j,i ”)-(t 1,j -t j ')
wherein T is j,i Representing the time of flight of the ith backscatter echo pulse of the jth spectral channel after time domain correction; t is t 2，j,i Representing the instant t of the recorded ith backscattered echo pulse of the jth spectral channel j,i "represents the propagation time error, t, of the ith backscattered echo pulse of the jth spectral channel 1，j Representing the recorded moment of the pulse emitted by the jth spectral channel, t j ' represents the propagation time error of the j-th spectral channel transmit pulse,
the step 3 comprises the following steps:
step 3.1, calculating a time-domain corrected channel data multi-echo quality value MEQ based on the following formula j ，
In f (x) j Represents time-domain corrected channel data including transmit pulse and back-scattered echo pulse data, m represents waveform length, S noise,j Representing the mean square error of background noise of the jth spectral channel;
step 3.2, establishing a multi-channel weighted accumulation model f a (x) The weighted accumulated channel data ch3_index (j) is output,
wherein omega is j A weight representing the jth spectral channel; n represents the number of spectral channels participating in the accumulation; fa (x) is a multi-channel weighted accumulation model; f (x) j Representing time domain corrected channel data, including transmit pulse and backscattered echo pulse data,
the step 4 comprises the following steps:
step 4.1, establishing a Gaussian mixture fitting model;
step 4.2, extracting a to-be-optimized transmitting pulse parameter and a to-be-optimized back scattering echo pulse parameter from the weighted and accumulated channel data CH3_index (j), and respectively optimizing the to-be-optimized transmitting pulse parameter and the to-be-optimized back scattering echo pulse parameter according to a Gaussian mixture fitting model to obtain an optimal transmitting pulse parameter and an optimal back scattering echo pulse parameter;
step 4.3, calculating the distance between the central position of the optimal back scattering echo pulse and the central position of the corresponding transmitting pulse according to the optimal transmitting pulse parameter and the optimal back scattering echo pulse parameter,
the step 4.2 comprises the following steps:
step 4.2.1 searching the weighted and accumulated channel data CH3_index (j) for the maximum value ψ of the transmitted pulse intensity for each spectral channel 0,j The method comprises the steps of carrying out a first treatment on the surface of the Obtaining the maximum value psi of the emission pulse of each spectral channel 0,j Corresponding center position X 0,j The method comprises the steps of carrying out a first treatment on the surface of the Obtaining the maximum value psi of the intensity of the transmitted pulse 0,j Inflection point spacing F on left and right sides 0,j Maximum value psi of emission pulse intensity of each spectral channel 0,j Center position X 0,j Inflection point distance F 0,j Forming a group of to-be-optimized transmitting pulse parameters;
step 4.2.2, sequentially inputting each group of emission pulse parameters to be optimized into a Gaussian mixture fitting model, and obtaining the optimal emission pulse parameters corresponding to each spectrum channel by adopting an LM algorithm based on nonlinear least squares curve fitting, wherein the optimal emission pulse parameters comprise a maximum value psi 'of the optimal emission pulse intensity' 0,j Optimum center position X' 0,j Inflection point spacing F' 0,j ；
Step 4.2.3 searching the weighted accumulated channel data ch3_index (j) for the maximum value ψ of the respective backscattered echo pulse intensities for each spectral channel j，i The method comprises the steps of carrying out a first treatment on the surface of the Obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Corresponding center position X j,i The method comprises the steps of carrying out a first treatment on the surface of the Obtaining maxima ψ of individual backscattered echo pulse intensities for each spectral channel j，i Inflection point spacing F on left and right sides j,i The maximum value psi of the back scattering echo pulse intensity corresponding to each back scattering echo pulse corresponding to the spectrum channel j，i Center position X j,i Inflection point distance F j,i Forming a group of back scattering echo pulse parameters to be optimized;
step 4.2.4, synchronously traversing each spectrum channel in a multithread way, inputting each group of backscattering echo pulse parameters to be optimized in the spectrum channel into a mixed Gaussian fitting model, adopting an LM algorithm based on nonlinear least squares curve fitting to obtain each group of optimized backscattering echo pulse parameters corresponding to the spectrum channel, and in each group of optimized backscattering echo pulse parameters corresponding to the spectrum channel, enabling the amplitude of the backscattering echo pulse to be smaller than a threshold thr j Deleting the corresponding optimized backscattering echo pulse parameters to obtain the optimized backscattering echo pulse parameters including the maximum value of the backscattering echo pulse intensityψ' j，p Center position X' j,p Inflection point spacing F' j,p ，
The Gaussian mixture fitting model is as follows:
wherein f j (t) is a mixed Gaussian fitting model of the jth spectrum channel, and ψ represents the amplitude; t represents a sampling time; x represents a central position; f represents half width; k represents the number of backscattered echo pulses of the spectral channel; i denotes the sequence number of the backscattered echo pulse.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210325224.4A CN114707595B (en) | 2022-03-29 | 2022-03-29 | Spark-based hyperspectral laser radar multichannel weighting system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210325224.4A CN114707595B (en) | 2022-03-29 | 2022-03-29 | Spark-based hyperspectral laser radar multichannel weighting system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114707595A CN114707595A (en) | 2022-07-05 |
| CN114707595B true CN114707595B (en) | 2024-01-16 |
Family
ID=82171326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210325224.4A Active CN114707595B (en) | 2022-03-29 | 2022-03-29 | Spark-based hyperspectral laser radar multichannel weighting system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707595B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116049641B (en) * | 2023-04-03 | 2023-06-30 | 中国科学院光电技术研究所 | Point target feature extraction method based on infrared spectrum |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110794424A (en) * | 2019-11-13 | 2020-02-14 | 湖北大学 | Full-waveform airborne laser radar ground feature classification method and system based on feature selection |
| CN110794387A (en) * | 2019-11-28 | 2020-02-14 | 中国科学院合肥物质科学研究院 | Radiation calibration method of airborne hyperspectral imaging laser radar system |
| CN110927735A (en) * | 2019-11-21 | 2020-03-27 | 中国科学院武汉物理与数学研究所 | Multi-target distance measuring method based on multi-channel full-waveform laser radar data |
| CN111898662A (en) * | 2020-07-20 | 2020-11-06 | 北京理工大学 | Coastal wetland deep learning classification method, device, equipment and storage medium |
| CN111948669A (en) * | 2020-08-11 | 2020-11-17 | 锐驰智光(苏州)科技有限公司 | Hyperspectral data information acquisition system based on laser radar |
| CN113870131A (en) * | 2021-09-16 | 2021-12-31 | 中国科学院精密测量科学与技术创新研究院 | Highlight removal method for multispectral laser radar color point cloud |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9292747B2 (en) * | 2013-03-15 | 2016-03-22 | The Boeing Company | Methods and systems for automatic and semi-automatic geometric and geographic feature extraction |
| US10621438B2 (en) * | 2017-09-07 | 2020-04-14 | Syso OU | Hybrid hyperspectral augmented reality device |
| CN112180392B (en) * | 2019-07-02 | 2024-05-17 | 中国科学技术大学 | Atmospheric component detection laser radar based on dispersion gating |
-
2022
- 2022-03-29 CN CN202210325224.4A patent/CN114707595B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110794424A (en) * | 2019-11-13 | 2020-02-14 | 湖北大学 | Full-waveform airborne laser radar ground feature classification method and system based on feature selection |
| CN110927735A (en) * | 2019-11-21 | 2020-03-27 | 中国科学院武汉物理与数学研究所 | Multi-target distance measuring method based on multi-channel full-waveform laser radar data |
| CN110794387A (en) * | 2019-11-28 | 2020-02-14 | 中国科学院合肥物质科学研究院 | Radiation calibration method of airborne hyperspectral imaging laser radar system |
| CN111898662A (en) * | 2020-07-20 | 2020-11-06 | 北京理工大学 | Coastal wetland deep learning classification method, device, equipment and storage medium |
| CN111948669A (en) * | 2020-08-11 | 2020-11-17 | 锐驰智光(苏州)科技有限公司 | Hyperspectral data information acquisition system based on laser radar |
| CN113870131A (en) * | 2021-09-16 | 2021-12-31 | 中国科学院精密测量科学与技术创新研究院 | Highlight removal method for multispectral laser radar color point cloud |
Non-Patent Citations (3)
| Title |
|---|
| Multichannel Interconnection Decomposition for Hyperspectral LiDAR Waveforms Detected From Over 500 m;Binhui Wang et al.;《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》;第1-14页 * |
| Target Classification of Similar Spatial Characteristics in Complex Urban Areas by Using Multispectral LiDAR;Shalei Song et al.;《remote sensing》;第1-22页 * |
| 基于高维特征分析的高分辨率遥感影像 变化检测方法研究;张晨晓;《中国博士学位论文全文数据库基础科学辑》;第1-108页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114707595A (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hovi et al. | LiDAR waveform features for tree species classification and their sensitivity to tree-and acquisition related parameters | |
| CN105488770B (en) | A kind of airborne laser radar point cloud filtering method of object-oriented | |
| EP2846173B1 (en) | Ambiguity compensation in time-of-flight ranging | |
| Shao et al. | A 91-channel hyperspectral LiDAR for coal/rock classification | |
| CN101839981A (en) | Method and device for acquiring laser imaging echo waveform and level characteristics | |
| CN114707595B (en) | Spark-based hyperspectral laser radar multichannel weighting system and method | |
| CN112147601B (en) | Sea surface small target detection method based on random forest | |
| CN107703503B (en) | Point trace condensation method based on GPU acceleration | |
| CN110927735B (en) | Multi-target distance measuring method based on multi-channel full-waveform laser radar data | |
| Cain et al. | Convolutional neural networks for radar emitter classification | |
| CN106066154A (en) | A kind of target being applicable to quickly scan scene and the extracting method at control point thereof | |
| CN113408328B (en) | Gesture segmentation and recognition algorithm based on millimeter wave radar | |
| Yovel et al. | What a plant sounds like: the statistics of vegetation echoes as received by echolocating bats | |
| CN106772299B (en) | One kind is based on apart from matched PD radar weak target Dynamic Programming detection method | |
| CN116184394A (en) | Millimeter wave radar gesture recognition method and system based on multi-domain spectrogram and multi-resolution fusion | |
| CN112327265A (en) | Division and treatment detection method based on semantic segmentation network | |
| Zhou et al. | Land cover classification from full-waveform lidar data based on support vector machines | |
| WO2022000333A1 (en) | Radar detection method and related device | |
| CN105911537B (en) | A method of reducing active sonar detection blind area | |
| CN105890548B (en) | A kind of method using hole estimation forest parameters between trunk | |
| WO2023019573A1 (en) | Ranging method, waveform detection method, apparatus, and related device | |
| Kuc | Transforming echoes into pseudo-action potentials for classifying plants | |
| Korpela et al. | Airborne dual-wavelength waveform LiDAR improves species classification accuracy of boreal broadleaved and coniferous trees | |
| CN107202993A (en) | The big visual field laser three-dimensional imaging system of cascade acousto-optic sampled based on Full wave shape | |
| CN108007866B (en) | Fruit rheological parameter detection method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |